I don't think it has yet become sufficiently clear what this relatively simple encryption, which is based on 15th-century cipher systems, actually means.
If entire consonant clusters are treated as single units – which is largely unproblematic in Bavarian, since such clusters are already used flexibly and variably in the spoken language – this alone has a noticeable effect on entropy. This effect increases further when articles and other function words are compressed into a small number of prefix clitics (as in the case of qo and o): the predictability of symbol sequences necessarily rises sharply.
If this is combined with vowel neutralisation (e.g. u/o/e → o), a vowel shift in which e is displaced to word-final y, and with other symbols no longer representing single sounds but entire letter or sound groups, the system as a whole begins to appear highly regular and strongly constrained from the outside – regardless of how “normal” the underlying target text may be.
An important point here is that spoken Bavarian and related Upper German dialects already show a strong tendency towards reduction (apocope, syncope, cliticisation). That means that if the underlying text – the input to the cipher, so to speak – already produces many short, highly compressed word forms (often very short, sometimes even monosyllabic), this further amplifies the effects described above.
If, in addition, attached articles are concealed at the surface level by a simple and mechanically explainable rule (for example: if a preceding qo would make a word too long, the word is split in order to prevent qo from being immediately recognisable as an article), the word-length distribution shifts even further towards the profile observed in the Voynich Manuscript.
Conclusion:
It follows that low entropy and short word lengths (with positive autocorrelation) are not sufficient arguments for excluding a European language as the input of the Voynich text. They are fully compatible with a deterministic production system that strongly compresses and formally constrains a European (Bavarian) target text.
If entire consonant clusters are treated as single units – which is largely unproblematic in Bavarian, since such clusters are already used flexibly and variably in the spoken language – this alone has a noticeable effect on entropy. This effect increases further when articles and other function words are compressed into a small number of prefix clitics (as in the case of qo and o): the predictability of symbol sequences necessarily rises sharply.
If this is combined with vowel neutralisation (e.g. u/o/e → o), a vowel shift in which e is displaced to word-final y, and with other symbols no longer representing single sounds but entire letter or sound groups, the system as a whole begins to appear highly regular and strongly constrained from the outside – regardless of how “normal” the underlying target text may be.
An important point here is that spoken Bavarian and related Upper German dialects already show a strong tendency towards reduction (apocope, syncope, cliticisation). That means that if the underlying text – the input to the cipher, so to speak – already produces many short, highly compressed word forms (often very short, sometimes even monosyllabic), this further amplifies the effects described above.
If, in addition, attached articles are concealed at the surface level by a simple and mechanically explainable rule (for example: if a preceding qo would make a word too long, the word is split in order to prevent qo from being immediately recognisable as an article), the word-length distribution shifts even further towards the profile observed in the Voynich Manuscript.
Conclusion:
It follows that low entropy and short word lengths (with positive autocorrelation) are not sufficient arguments for excluding a European language as the input of the Voynich text. They are fully compatible with a deterministic production system that strongly compresses and formally constrains a European (Bavarian) target text.